GitHub - otto-torino/tazebao: A web application which provides a newsletter service to other applications

Otto srl newsletter web application.

What?

Tazebao is a web application which aims to provide a newsletter service to other sites/applications.

Every platform User can be associated to one Client and manage a newsletter. The platform provides an administrative area where the client user can manage its data.

Tazebao provides the following entities:

subscriber
list of subscribers
topic
template
campaign
planning
dispatch
dispatch queue log
tracking

Each Subscriber must belong to one or more lists (SubscriberList). Subscriber's e-mail must be unique for each client. Each Campaign must belong to a Topic, which stores the information about the sender name and address. Tazebao integrates mosaico, a responsive e-mail template generator. The generated template can be included in the campaign. The newsletter is sent creating a Dispatch, where the user chooses one or more lists of subscribers which will receive the e-mail. Newsletter can be scheduled, planning the effective dispatch. If the html template of the email contains the closing body tag, then a tracking code is added automatically in order to track if the user opens the e-mail. These statistics are inaccurate, because the tracking code works only if the user loads the e-mail images. Tazebao provides a link templatetag which can be used to track user clicks (the destination url is transformed in a Tazebao url, which executes a view which detects the event and redirects the user to the original url).

The Subscriber has an info text field which can be used to store additional information, for example in a json format.

E-mails are prepared asynchronously through a celery task, queued and sent in blocks. The scheduling must be set with a crontab, Tazebao uses the django-mail-queue application in order to manage this stuff.
Just set a crontab which calls

$ python manage.py send_queued_messages

Tazebao provides a REST webservice in order to retrieve, create, edit and delete subscribers and lists. Each Client can then implement its own registration form and use the provided API to update Tazebao DB. Also the unsubscription feature must be implemented by the Client, which will then use the API to delete the record from Tazebao. Tazebao provides a functionality which can be used to create a signed string of subscriber's data used to generate the unsubscription url. Unsubscription text is managed in the Topic section of Tazebao. For example you can generate a URL which includes a signature of the ID and EMAIL of the subscriber, then you can check the signature in your own application, and if matches call the delete action of the API.

Setup

As superuser you can create clients, which are users that receive a set of keys necessary for communicating with the REST Web Service. You should also create a Client's Group, associated with each Client user, with the following permissions:

mosaico | template | Can add template
mosaico | template | Can change template
mosaico | template | Can delete template
mosaico | upload | Can add upload
mosaico | upload | Can change upload
mosaico | upload | Can delete upload
newsletter | Campagna | Can add Campagna
newsletter | Campagna | Can change Campagna
newsletter | Campagna | Can delete Campagna
newsletter | Client | Can change Client
newsletter | Invio | Can change Invio
newsletter | Invio | Can delete Invio
newsletter | Iscritto | Can add Iscritto
newsletter | Iscritto | Can change Iscritto
newsletter | Iscritto | Can delete Iscritto
newsletter | Lista iscritti | Can add Lista iscritti
newsletter | Lista iscritti | Can change Lista iscritti
newsletter | Lista iscritti | Can delete Lista iscritti
newsletter | Planning | Can add Planning
newsletter | Planning | Can change Planning
newsletter | Planning | Can delete Planning
newsletter | Topic | Can add Topic
newsletter | Topic | Can change Topic
newsletter | Topic | Can delete Topic
newsletter | Tracking | Can change Tracking
newsletter | Tracking | Can delete Tracking
newsletter | Log coda di invio | Can change Log coda di invio
newsletter | Log coda di invio | Can delete Log coda di invio

so that clients can manage their account through the admin interface.

REST Web Service

The admin application authentication is a session based authentication, while the web service uses HMAC (Keyed-Hash Message Authentication Code).

Each Client gets an ID_KEY and a SECRET_KEY through which authenticate to Tazebao. The Signature must be passed in the HTTP header of every request, Tazebao uses django-rest-framework-httpsignature to manage this stuff.

REST Service Authentication

In order to perform valid requests to the REST resources, every request must be signed in the HTTP header. The signature must be calculated this way:

SIGNATURE = Base64(Hmac('SECRET_KEY', "date: DATETIME", 'sha256'))

And the request should be:

~$ curl -v -H 'Date: "DATETIME"' -H 'Authorization: Signature keyId="ID_KEY",algorithm="hmac-sha256",headers="date",signature="SIGNATURE"'

this was descriptive code, take a look at the client repository to see how to implement this stuff in real languages.

REST resources

Tazebao provides a REST Web Service which allows to manage subscribers and lists data.

Subscribers Lists

This section describes possible CRUD requests to manage subscribers's lists.

Retrieve lists

GET http://localhost:8000/api/v1/newsletter/subscriberlist/

returns a list of subscribers' lists associated to the authenticated client, i.e.:

[{"id":1,"name":"Journalists"},{"id":8,"name":"ACABS"}]

Get a list

GET http://localhost:8000/api/v1/newsletter/subscriberlist/<LIST_ID>/

returns the list object represented by LIST_ID, i.e.:

{"id":1,"name":"Journalists"}

Add a list

POST http://localhost:8000/api/v1/newsletter/subscriberlist/

the POST data should be a json containing the name field, i.e.:

"{"name":"Journalists"}"

for example with PHP:

$post_data = json_encode(array('name' => 'Journalists'))

Edit a list

PUT http://localhost:8000/api/v1/newsletter/subscriberlist/<LIST_ID>/

the submitted data should be a json containing the name field, i.e.:

"{"name":"Journalists"}"

Delete a list

DELETE http://localhost:8000/api/v1/newsletter/subscriberlist/<LIST_ID>/

Subscribers

This section describes possible CRUD requests to manage subscribers.

Retrieve subscribers

GET http://localhost:8000/api/v1/newsletter/subscriber/

returns a paginated (200 for page) list of subscribers associated to the authenticated client, i.e.:

{
    "count": 300,
    "next": "http://localhost:8000/api/v1/newsletter/subscriber/?page=2",
    "previous": null,
    "results": [
        {
            "id":1,
            "client":1,
            "email":"[email protected]",
            "subscription_datetime":"2016-09-08T14:49:58.860948Z",
            "info":"custom stuff",
            "lists":[1]
        },
        {
            "id":2,
            "client":1,
            "email":"[email protected]",
            "subscription_datetime":"2016-09-08T14:59:58.860948Z",
            "info":"",
            "lists":[1, 2]
        },
        // ... 198 more
    ]
}

Get a subscriber

GET http://localhost:8000/api/v1/newsletter/subscriber/<SUBSCRIBER_ID>/

returns the subscriber object represented by SUBSCRIBER_ID, i.e.:

{
    "id":2,
    "client":1,
    "email":"[email protected]",
    "subscription_datetime":"2016-09-08T14:59:58.860948Z",
    "info":"",
    "lists":[1, 2]
}

Add a subscriber

POST http://localhost:8000/api/v1/newsletter/subscriber/

the POST data should be a json of the following format:

"{"email":"[email protected]","info":"\"firstname=\"meow\"\"","lists":["1"]}"

email and lists fields are required. If you try to add a subscriber with an email address already associated to the client, you get a 400 response describing the error.

Edit a subscriber

PUT http://localhost:8000/api/v1/newsletter/subscriber/<SUBSCRIBER_ID>/

the POST data should be a json of the following format:

"{"email":"[email protected]","info":"\"firstname=\"meow\"\"","lists":["1"]}"

email and lists fields are required.

Delete a subscriber

DELETE http://localhost:8000/api/v1/newsletter/subscriber/<SUBSCRIBER_ID>/

Campaigns

Campaign models are readonly, so you can only retrieve a list of paginated campaigns, or a single campaign. This resource is usefull in order to display the newsletter content in your own application, or to create a newsletter archive.

Retrieve campaigns

GET http://localhost:8000/api/v1/newsletter/campaign/

returns a paginated (20 for page) list of campaigns associated to the authenticated client, i.e.:

{
    "count": 2,
    "next": "http://localhost:8000/api/v1/newsletter/campaign/?page=2",
    "previous": null,
    "results": [
        {
            "id": 1,
            "topic_id": 1,
            "topic": "New Articles",
            "name": "Let's talk about birds!",
            "insertion_datetime": "2016-09-08T14:51:06.980733Z",
            "last_edit_datetime": "2016-09-22T11:42:30.398038Z",
            "subject": "A new article on birds.com!",
            "plain_text": "plain bla bla",
            "html_text": "html bla bla",
            "view_online": true,
            "url": "http://localhost:8000/newsletter/client-slug/2016/09/08/campaign-slug/"
        }
        // ... 19 more
    ]
}

additional filters can be used to retrieve campaigns:

date_from in the format YYYY-MM-DD
date_to in the format YYYY-MM-DD
subject
text

just add them in the query string, i.e.

http://localhost:8000/api/v1/newsletter/campaign/?subject=foo&date_from=2018-09-12

Get a campaign

GET http://localhost:8000/api/v1/newsletter/campaign/<CAMPAIGN_ID>/

returns the campaign object represented by CAMPAIGN_ID, i.e.:

{
    "id": 1,
    "topic_id": 1,
    "topic": "New Articles",
    "name": "Let's talk about birds!",
    "insertion_datetime": "2016-09-08T14:51:06.980733Z",
    "last_edit_datetime": "2016-09-22T11:42:30.398038Z",
    "subject": "A new article on birds.com!",
    "plain_text": "plain bla bla",
    "html_text": "html bla bla",
    "view_online": true,
    "url": "http://localhost:8000/newsletter/client-slug/2016/09/08/campaign-slug/"
}

Dispatches

Dispatch models are readonly, so you can only retrieve a list of paginated dispatches, or a single dispatch.

Retrieve dispatches

GET http://localhost:8000/api/v1/newsletter/dispatch/

returns a paginated (20 for page) list of dispatches associated to the authenticated client, i.e.:

{
    "count": 2,
    "next": "http://localhost:8000/api/v1/newsletter/campaign/?page=2",
    "previous": null,
    "results": [
        {
            "id": 800,
            "campaign": 387,
            "lists": [
                23
            ],
            "started_at": "2019-06-06T10:28:34+02:00",
            "finished_at": "2019-06-06T10:28:34+02:00",
            "error": false,
            "error_message": null,
            "success": true,
            "open_statistics": true,
            "click_statistics": false,
            "sent": 1,
            "error_recipients": ""
            "open_rate": 100,
            "click_rate": 100,
            "trackings": [
                {
                    "id": 159589,
                    "datetime": "2019-06-06T16:46:40+02:00",
                    "type": "apertura",
                    "subscriber_email": "[email protected]",
                    "notes": null
                },
                {
                    "id": 159590,
                    "datetime": "2019-06-06T17:50:54+02:00",
                    "type": "click",
                    "subscriber_email": "[email protected]",
                    "notes": "http:\/\/www.google.it"
                }
            ],
            "bounces": [
                {
                    "id": 1,
                    "datetime": "2019-10-18T15:40:19+02:00",
                    "dispatch": 800,
                    "from_email": "[email protected]",
                    "subscriber_email": "[email protected]",
                    "message": "host gmail-smtp-in.l.google.com[108.177.15.26] said: 550-5.1.1 The email account that you tried to reach does not exist. Please try 550-5.1.1 double-checking the recipient's email address for typos or 550-5.1.1 unnecessary spaces. Learn more at 550 5.1.1 https:\/\/support.google.com\/mail\/?p=NoSuchUser f2si4669900wrx.476 - gsmtp (in reply to RCPT TO command)",
                    "status": "bounced"
                }
            ]
        },
        // ... 19 more
    ]
}

additional filters can be used to retrieve campaigns:

date_from in the format YYYY-MM-DD
date_to in the format YYYY-MM-DD
campaign

just add them in the query string, i.e.

http://localhost:8000/api/v1/newsletter/dispatch/?subject=foo&date_from=2018-09-12

Get a dispatch

GET http://localhost:8000/api/v1/newsletter/dispatch/<DISPATCH_ID>/

returns the dispatch object represented by DISPATCH_ID, i.e.:

Mosaico Integration

Tazebao relies on mosaico in order to provide a template generator engine. The mosaico core lib has been touched only to prevent url encoding of django template tags and variables. The e-mail templates have been hacked in order to integrate some Tazebao functionalities.

Getting Started

To be up and running for local development just follow these steps:

clone the repository
$ git clone https://github.com/otto-torino/tazebao.git
cd the new project
$ cd [repo_name]
create a virtualenv
$ virtualenv --no-site-packages .virtualenv
activate it
$ source .virtualenv/bin/activate
install requirements
$ pip install -r [repo_name]/requirements/local.txt
create a .env file
$ touch .env
config environment
$ dotenv set DJANGO_SETTINGS_MODULE core.settings.local
$ dotenv set DB_PASSWORD <whatever> $ dotenv set SECRET_KEY <whatever>
run the server
$ bin/runserver
enjoy
http://localhost:8000

E-mails are queued and sent in blocks. The scheduling is managed with crontab, but the queue preparation is managed asynchronously with a celery task. That's because preparing thousand of e-mails for sending is a dispendious job to do within a request, and it would require a not-pleasant timeout setting.

To start celery services:

supervisord -c supervisord.conf.local

To monitor celery workers:

supervisorctl -c supervisord.conf.local
inside the prompt start, stop and see the status:
- start all
- stop all
- status

Remote setup

Remote setup is done with ansible, using the root user, the scripts are configured for debian based distros.
First you need to configure your remote host inventory and variables.

Create a file provisioning/ansible_remote_inventory with the following content:

remote ansible_ssh_host=<YOUR_DOMAIN_HERE>

then create a provisioning/ansible_remote_variables file:

---
provisioning: 'remote'
domain: '<YOUR_DOMAIN_HERE>'
mysql_root_password: '<MYSQL_ROOT_PWD>'
db_user: '<CHOOSE_A_DB_USER>'
db_password: '<CHOOSE_A_DB_PWD>'
db_name: '<CHOOSE_A_DB_NAME>'
user: '<CHOOSE_THE_REMOTE_USER_TIED_TO_THE_APP>'
password: '<..AND_ITS_PWD>'
webapp_dir: '<REMOTE_WEBAPP_DIR>'

Ok you're ready for the provisioning phase, launch

$ bin/ansible_remote

and provide the remote root password when prompted.

If all goes well, now you should have your remote machine ready for deploy. Visit your domain and you should see a maintenance page already there.

Celery configuration

Celery has a production ready configuration file you find in the ROOT of the repository:

supervisord.conf.production

You need to customize it in order to match your remote user and rempte paths, just edit the following section:

[program:tazebao-workers]
command=/home/tazebao/www/tazebao/.virtualenv/bin/celery --app=core.celery:app worker --loglevel=INFO
directory=/home/tazebao/www/tazebao/releases/current
user=tazebao
numprocs=1
stdout_logfile=/home/tazebao/www/tazebao/logs/celery-beat.log
stderr_logfile=/home/tazebao/www/tazebao/logs/celery-beat.log
autostart=true
autorestart=true
startsecs=10

(in this case the user is tazebao and the webapp remote dir is /home/tazebao/www/tazebao, leave the nested parts of the paths as they are, or it'll not work).

As for now it is not uploaded during the provisioning phase with ansible, so you have to do it manually:

$ scp supervisord.conf.production USER@HOST:/PATH/TO/REMOTE/WEBAPP/DIR

Troubleshooting

SSH connection error

If you see this error

fatal: [remote] => Using a SSH password instead of a key is not possible because Host Key checking is enabled and sshpass does not support this.  Please add this host's fingerprint to your known_hosts file to manage this host.

means you missed adding an entry for one or more hosts in the ~/.ssh/known_hosts file, it's enough to

DB task error

If an error occurs in the create db user task:

msg: (1396, "Operation CREATE USER failed for ...

maybe you're trying to create a user which already existed and was delete. In this case just run this sql in the mysql database:

FLUSH PRIVILEGES;

see this answer on stack overflow

Deploy and Stuff

In the deployment process, the last revision of the git repository is deployed to the remote server. So be sure to have committed all your changes:

$ git add --all
$ git commit -a -m "first commit"

Be sure the provided remote user has ssh access to the remote host, then deploy should be as easy as:

$ cd [repo_name]
$ fab production deploy

launched inside the root/repo_name folder. This command does the following things:

create an archive of the last repository revision
upload it to the server
untar it in a folder "app-revision_id" inside the releases folder
copy the .env file inside this folder
upgrade the virtualenv
collectstatic
migrations
move the current release in the previous release (releases/previous)
link the releases/current folder to the new release folder
restart uwsgi and nginx
open a shell in the remote

When performing the first deploy you can create a superuser account using the shell which the script leaves open at the end.

When deploying the first time you need also to start celery workers, so inside the open shell session:

$ supervisord -c ../../supervisord.conf.production

As in locale you can monitor the supervisor status and start and stop processes:

$ supervisorctl -c ../../supervisord.conf.production

When you deploy again it won't be necessary to repeat the command above, because the processes are already running. But if you changed the configuration or the tasks you may need to restart this stuff, so:

$ supervisorctl -c ../../supervisord.conf.production
$ > stop all
$ > exit
$ supervisord -c ../../supervisord.conf.production

Then just check with ps ax | grep supervisor and ps ax | grep celery if all is ok, if not just kill -9 the processes and restart supervisord.

Other useful fab commands

rollback

$ fab production rollback

If the deploy revision is broken, or introduces unexpected errors, with this command it is possible to rollback to the previous revision. Launching it another time will swap between the two revisions.

restart_uwsgi

$ fab production restart_uwsgi

Restarts the uwsgi service

restart_server

$ fab production restart_server

Restarts the web server

restart

$ fab production restart

Restarts the uwsgi service and the web server

dump_db_snapshot

$ fab production dump_db_snapshot

Downloads the production current db snapshot in the backup folders. The dumped file has the remote current revision name.

Requires the remote db user password.

load_db_snapshot

$ fab production load_db_snapshot

Loads the current remote db snapshot in the local db.